A Statistically Efficient Method for Fusion of Multimodal Images in NSCT Domain

Image fusion is a procedure that enhances the understanding of a scene by integrating information collected from various imaging sensors. This technique is extensively applied in the field of image processing due to its effectiveness. Multifocus image fusion is a technique that deals with the challenge of combining two or more images captured of the same scene but focused on different depths. In medical image fusion, the primary benefit is the ability to extract relevant data from input images captured by different devices for the same organ, which is very useful for disease diagnosis. This concept of image processing holds significant importance across various applications as it contributes to improved perception, decision-making, and overall system performance.In the present study, we propose an efficient algorithm based on the non-subsampled contourlet transform (NSCT). This method takes into consideration the strong linear dependency between high-frequency coefficients in the input images. The low-frequency coefficients are fused using different fusion rules based on principal component analysis and a salience match measure. We conducted experiments using three sets of multimodal images, and the performance of our algorithm is compared to various spatial and transform domain image fusion methods presented in existing literature. We employed quantitative evaluation techniques based on diverse image fusion quality criteria to assess the fusion results. Additionally, we conducted subjective assessments to gauge the performance of the resulting fused images.